Closed-Loop Autonomous Experimentation in Materials Discovery

Abstract

We report a materials discovery platform sustaining a fully closed experimental loop for one month without human intervention. Candidate yield improves over the run, indicating directed search rather than enumeration. We are explicit that the research question was framed by the human team, and discuss what would be required for that step to be automated.

Key findings

• 01Full loop closed without intervention for one month
• 02Candidate yield improves over the run
• 03Throughput bounded by physical experiment, not by proposal generation
• 04Research question framing remained human throughout

Limitations

Published at equal prominence to the findings. A paper’s limitations are usually the part that determines how much its result should move your beliefs.

• Single narrow materials domain
• Human-framed research question
• Comparison against a human-directed baseline is not controlled